Machine learning system driven by high-performance material and application method thereof
By adopting integrated machine learning systems in the research and development of high-performance materials, including technologies such as data cleaning, feature extraction, cross-verification and deep learning networks, the problems of imperfect data processing and lack of systematicity in hyperparameter selection are solved, and high performance is achieved Accurate prediction and optimization of materials improve R&D efficiency and application scope.
Patent Information
- Application Number
- CN202411211890.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-30
- Publication Date
- 2025-01-07
AI Technical Summary
The prior art has problems such as imperfect data processing, lack of systematic hyperparameter selection, and limited application range of prediction results in the research and development of high-performance materials, resulting in insufficient accuracy and low efficiency in material performance prediction.
Using high-performance materials-driven machine learning systems, including data processing modules, machine learning model training modules, prediction modules and feedback optimization modules, through data cleaning, feature extraction, cross-validation, deep learning networks, gradient descent method and genetic algorithms. , optimize the model training and prediction process.
It improves the accuracy and efficiency of material performance prediction, ensures data processing quality, expands the scope of application, and achieves accurate prediction and optimization of high-performance materials in specific fields.
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of performance prediction of high-performance materials, and in particular to a machine learning system driven by high-performance materials and an application method thereof. Background Art
[0002] With the continuous advancement of science and technology, high-performance materials are increasingly used in aviation, aerospace, electronics, energy and other fields. These materials usually have excellent physical, chemical or mechanical properties, which are of great significance in promoting technological innovation in related fields. However, the research and development of high-performance materials often requires a lot of experiments and tests, which is long, costly, and difficult to ensure the stability and reliability of material performance.
[0003] In recent years, machine learning technology has received extensive attention in the field of materials science, especially showing great potential in material performance prediction, structural optimization, fault diagnosis, etc. Traditional material performance prediction methods mainly rely on physical models and empirical formulas, which have certain limitations when dealing with complex systems. In contrast, machine learning technology can learn rules from a large amount of experimental data and achieve rapid and accurate prediction of material properties.
[0004] Although existing machine learning techniques have achieved certain results in material property prediction, the following problems still exist:
[0005] 1. The data processing method is not perfect, and it is difficult to effectively extract key features that help improve the model's predictive ability;
[0006] 2. During the training of machine learning models, the selection and optimization of hyperparameters mainly rely on experience and lack a systematic approach;
[0007] 3. The application scope of the prediction results is limited, and it is difficult to accurately predict and optimize high-performance materials in specific fields.
[0008] In response to the above problems, the present invention proposes a high-performance material-driven machine learning system and its application method, aiming to improve the accuracy and efficiency of material performance prediction and provide strong support for the research and development and application of high-performance materials. Summary of the invention
[0009] The present invention is achieved through the following technical solutions:
[0010] High-performance materials-driven machine learning systems, including:
[0011] A data processing module for acquiring and preprocessing data related to the high-performance material, wherein the data at least includes material composition, structural parameters, temperature changes, stress-strain relationships, etc., and the data processing module also includes:
[0012] A data cleaning unit, used to remove invalid data and outliers, wherein the outliers are detected using a Z-score method, with a threshold set to 3σ, i.e., three times the standard deviation;
[0013] A feature extraction unit, used to extract key features from the original data that help improve the prediction ability of the model, including but not limited to frequency domain features obtained by Fourier transform and dimensionality reduction features obtained by principal component analysis (PCA);
[0014] A machine learning model training module, used for training one or more machine learning models based on the preprocessed data, wherein the model training process adopts a cross-validation technique with a validation set ratio of 20%;
[0015] A prediction module, used to predict the performance of the new high-performance material using the trained model, wherein the prediction results include but are not limited to key performance indicators such as strength, toughness, and conductivity of the material;
[0016] A feedback optimization module is used to adjust the parameters of the machine learning model according to the prediction results to improve the prediction accuracy. The feedback optimization adopts the gradient descent method, and the learning rate is initially set to 0.01 and gradually decays as the number of iterations increases.
[0017] Furthermore, the machine learning model training module uses a deep learning network for model training, wherein the deep learning network includes but is not limited to a convolutional neural network (CNN) with at least 5 hidden layers or a recurrent neural network (RNN) with a long short-term memory unit (LSTM).
[0018] Furthermore, the activation function of the deep learning network adopts the ReLU function, the loss function adopts the mean square error (MSE), and the batch normalization technology is used in the model training process to accelerate convergence.
[0019] Furthermore, the feedback optimization module uses a genetic algorithm to automatically optimize model parameters, wherein the population size of the genetic algorithm is 50, the selection rate is 0.3, the crossover rate is 0.7, the mutation rate is 0.01, and the evolutionary generation is 100 generations.
[0020] Furthermore, the high-performance material is a new type of conductive material for power systems, and the conductive performance of the material is calculated by the following formula:
[0021] σ=f(composition, structural parameters, T, strain)
[0022] Among them, σ represents conductivity, T represents temperature, composition and structural parameters represent the basic properties of the material, strain represents the amount of strain received by the material, and f is the output function output by the machine learning model.
[0023] A method for applying a high-performance material-driven machine learning system comprises the following steps:
[0024] S1: Collect relevant data of the high-performance material, including but not limited to material composition, structural parameters, temperature changes, stress-strain relationship, etc.;
[0025] S2: preprocessing the data using the data processing module, including data cleaning and feature extraction;
[0026] S3: using the machine learning model training module to train the machine learning model, wherein the model training process adopts a cross-validation technique, the validation set ratio is 20%, and a genetic algorithm is used for hyperparameter optimization;
[0027] S4: using the prediction module to predict the performance of the new high-performance material, wherein the prediction results include but are not limited to key performance indicators such as strength, toughness, and conductivity of the material;
[0028] S5: Utilizing the feedback optimization module to optimize the machine learning model according to the prediction results, including but not limited to adjusting model parameters and hyperparameters to improve prediction accuracy.
[0029] Furthermore, the high-performance material is a new type of conductive material for use in power systems, and the conductive performance prediction results of the material are used to guide the design and improvement of the material to improve the overall performance of the power system.
[0030] Compared with the existing technology, the advantages of the present invention are:
[0031] 1. Perfect data processing method:
[0032] The present invention can more effectively remove invalid data and outliers, and extract key features from raw data that help improve the model's prediction capabilities by integrating a data cleaning unit and a feature extraction unit. This method improves the quality and efficiency of data processing and ensures a more solid data foundation for model training.
[0033] 2. Systematic model training and hyperparameter optimization:
[0034] The present invention adopts cross-validation technology to optimize the model training process, and optimizes hyperparameters through systematic methods such as genetic algorithms, thereby improving the objectivity and efficiency of model training.
[0035] 3. Wide range of applications and accurate predictions:
[0036] The present invention provides a specific application method for high-performance materials, which can accurately predict and optimize high-performance materials in specific fields such as new conductive materials. This not only expands the application scope of machine learning systems, but also improves the accuracy of predictions. DETAILED DESCRIPTION
[0037] Here, exemplary embodiments will be described in detail, and unless otherwise indicated, the embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present disclosure. Instead, they are only examples consistent with some aspects of the present disclosure as detailed in the appended claims.
[0038] The present invention provides a high-performance material-driven machine learning system and an application method thereof, and the specific implementation methods are as follows:
[0039] 1. Implementation of data processing module:
[0040] Data collection: First, collect data related to high-performance materials, including material composition, structural parameters, temperature changes, stress-strain relationships, etc.
[0041] Data cleaning: The data cleaning unit was used to detect outliers using the Z-score method, and the threshold was set to 3σ, i.e., three times the standard deviation, to remove invalid data and outliers.
[0042] Feature extraction: Using the feature extraction unit, Fourier transform is applied to convert the time domain data into the frequency domain, frequency domain features are extracted, and principal component analysis (PCA) is used to reduce the dimension of the data and extract the main features.
[0043] 2. Implementation of machine learning model training module:
[0044] Model selection: Choose a deep learning network such as a convolutional neural network (CNN) with at least 5 hidden layers or a recurrent neural network (RNN) with long short-term memory units (LSTM) for model training.
[0045] Model configuration: The activation function of the deep learning network uses the ReLU function, the loss function uses the mean square error (MSE), and batch normalization technology is used during training to accelerate model convergence.
[0046] Cross-validation: During the model training process, cross-validation technology is used, and the validation set ratio is set to 20% to ensure that the model has good generalization ability.
[0047] 3. Implementation of prediction module:
[0048] Performance prediction: Use the trained model to predict the performance of new high-performance materials, and output prediction results for key performance indicators including material strength, toughness, conductivity, etc.
[0049] 4. Implementation of feedback optimization module:
[0050] Parameter optimization: According to the prediction results, the gradient descent method is used to adjust the model parameters. The initial learning rate is set to 0.01 and gradually decays according to the number of iterations.
[0051] Genetic algorithm optimization: Genetic algorithm is used to automatically optimize model parameters, with the population size set to 50, the selection rate to 0.3, the crossover rate to 0.7, the mutation rate to 0.01, and the evolutionary generation to 100 generations.
[0052] 5. Specific steps of application method:
[0053] S1: Collect relevant data.
[0054] S2: Use the data processing module to perform data preprocessing.
[0055] S3: Use the machine learning model training module to train the model and combine it with the genetic algorithm for hyperparameter optimization.
[0056] S4: Use the prediction module to predict the performance of new materials.
[0057] S5: Use the feedback optimization module to optimize model parameters and hyperparameters based on the prediction results.
[0058] 6. Implementation of specific application scenarios:
[0059] The method of the present invention is used to predict the conductivity of new conductive materials for power systems, and the prediction results are used to guide the design and improvement of materials to improve the overall performance of the power system.
[0060] The above is only one of the specific implementations of the present invention, but the technical features of the present invention are not limited thereto. Any modification, equivalent replacement or improvement made within the technical concept of the present invention shall be included in the protection scope of the present invention.
Claims
1. A high-performance material-driven machine learning system, characterized by: include: A data processing module for acquiring and preprocessing data related to the high-performance material, wherein the data at least includes material composition, structural parameters, temperature changes, stress-strain relationships, etc., and the data processing module also includes: A data cleaning unit, used to remove invalid data and outliers, wherein the outliers are detected using a Z-score method, with a threshold set to 3σ, i.e., three times the standard deviation; A feature extraction unit, used to extract key features from the original data that help improve the prediction ability of the model, including but not limited to frequency domain features obtained by Fourier transform and dimensionality reduction features obtained by principal component analysis (PCA); A machine learning model training module, used for training one or more machine learning models based on the preprocessed data, wherein the model training process adopts a cross-validation technique with a validation set ratio of 20%; A prediction module, used to predict the performance of the new high-performance material using the trained model, wherein the prediction results include but are not limited to key performance indicators such as strength, toughness, and conductivity of the material; A feedback optimization module is used to adjust the parameters of the machine learning model according to the prediction results to improve the prediction accuracy. The feedback optimization adopts the gradient descent method, and the learning rate is initially set to 0.01 and gradually decays as the number of iterations increases.
2. The high performance material driven machine learning system according to claim 1, characterized in that: The machine learning model training module uses a deep learning network to perform model training, wherein the deep learning network includes but is not limited to a convolutional neural network (CNN) with at least 5 hidden layers or a recurrent neural network (RNN) with a long short-term memory unit (LSTM).
3. The high performance material driven machine learning system according to claim 2, characterized in that: The activation function of the deep learning network adopts the ReLU function, the loss function adopts the mean square error (MSE), and batch normalization technology is used in the model training process to accelerate convergence.
4. The high performance material driven machine learning system according to claim 1, characterized in that: The feedback optimization module uses a genetic algorithm to automatically optimize model parameters, wherein the population size of the genetic algorithm is 50, the selection rate is 0.3, the crossover rate is 0.7, the mutation rate is 0.01, and the evolutionary generation is 100 generations.
5. The high performance material driven machine learning system according to claim 1, characterized in that: The high-performance material is a new type of conductive material used in power systems. The conductive performance of the material is calculated by the following formula: σ=f(composition, structural parameters, T, strain) Among them, σ represents conductivity, T represents temperature, composition and structural parameters represent the basic properties of the material, strain represents the amount of strain received by the material, and f is the output function output by the machine learning model.
6. An application method of a machine learning system driven by high-performance materials according to any one of claims 1 to 5, characterized in that: The following steps are involved: S1: Collect relevant data of the high-performance material, including but not limited to material composition, structural parameters, temperature changes, stress-strain relationship, etc.; S2: preprocessing the data using the data processing module, including data cleaning and feature extraction; S3: using the machine learning model training module to train the machine learning model, wherein the model training process adopts a cross-validation technique, the validation set ratio is 20%, and a genetic algorithm is used for hyperparameter optimization; S4: using the prediction module to predict the performance of the new high-performance material, wherein the prediction results include but are not limited to key performance indicators such as strength, toughness, and conductivity of the material; S5: Utilizing the feedback optimization module to optimize the machine learning model according to the prediction results, including but not limited to adjusting model parameters and hyperparameters to improve prediction accuracy.
7. The application method according to claim 6, characterized in that: The high-performance material is a new type of conductive material used in power systems, and the conductive performance prediction results of the material are used to guide the design and improvement of the material to improve the overall performance of the power system.